利用参考图像的遥感图像模糊估计与恢复

《利用参考图像的遥感图像模糊估计与恢复》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、WANGTianhuieta1．：Blurestimationandrestorationofremotesensingimagesusingareferenceimage461(m，)P(YlYc；h，)～N(n310，)，(7)0rP(Yl；，)=e坤Fig．1Degradationmodelinimageacquisition[_]’㈣whereHisthedegradationfunctionintheformofcircularmat—Fromthemodel，wecangetthedegrad

2、ationequationrix．Takelogarithmoperationtobothsidesoftheaboveequa-Y=x圆h+／1，(1)tion，andwehavewhere0denotestheconvolutionoperationandlisusedins—一teadofnonlyfortheconvenienceoflateruse．专枷·Nowletbeablurfreeimageofthesamescene，i．e．，(y-臌)一mIno-2+ch0-1，thenthefol

3、lowingequationholds．2，(9)whereMistheimagesizeandCisaconstant．Wecangetano—=x0％+n2+2，(2)wherewealsoassume，tobeGaussianwhitenoiseofzerotherestimationexpressionbysubstituteEq．(6)toEq．(9)，i．e．，mean．盯gmin{(y城W_ethenobtainfromEq．(1)andEq．(2)thatY=0h+n3．(3)Noten3

4、isGaussiannoiseofzeromean，forGaussiannoise()+M}．(10)alwaysremainGaussianwhenitpassesalinearsystem．Butn3Theequationabovecanbeinducedtoalinearsystemaboutdoesnotnecessarilyremainwhite，whichfortunatelyhasnohandwhosesolutionisunique，ontheconditionthattheefecto

5、nourfollowingdiscussion．degradationfunctionissymmetricand≥0，∑f=landhhasWecanseefromEq．(3)thatthetaskofblindPSFestima-tionisconvertedtoadeconvolutionproblem，thankstoablur2．3Restorationfreeimageofthesamescene(norequirementonnoise)，thatis，theestimationofh，gi

6、venyand．AfterhavingfinishedthePSFestimation，wecanusetheAftertheestimationofhiscompleted，therestorationofTotalVariation(TV)regularizationmethod(Zou，2001)tores—becomesasimilardecOnvO1utiOntask，i．e．，toestimatex，knowntoretheoriginalimage．Thismethodcanbetterpr

7、eservetheYandh，accordingtoEq．(1)．detailinformationontheimageandsuppressthenoise．The2．2PSFestimationcostfunctionis，(／)=1Jrg一酬l+l+，(11)IfweestimatePSFwithregularizationmethod，wehavetodealwithaniterativeoptimizationprocedure，owingtothewherefandgarethefimctio

8、naloftheundistortedimageandrestrictionsonhsuchasnon—negative，normalizedandfinitetheobservednoisyandblurredimage，respectively,Kisthesupported(Yuaneta1．，2007)．Thatiswhatwetrytoavoid，bluroperator．Visthegradientoperator